Optical communications systems as presently contemplated typically use light sources and photodetectors that are optically coupled to each other by means of a glass transmission line which is commonly referred to as an optical fiber. The light source generally contemplated for use in such systems comprises a semiconductor laser diode which has an active layer, i.e., the region in which electron-hole recombination occurs, of relatively small dimensions perpendicular to the direction of radiation emission and which is generally in the form of a stripe. Optical communications systems and components thereof are discussed in Optical Devices & Fibers, 1982, edited by T. Suematsu.
For the high data rate communications systems contemplated, narrow spectral output corresponding to single mode operation of the light source is desirable as this minimizes the problems--e.g., limited repeater spacing, relatively low bit rate, etc. (for single channel as well as wavelength multiplexed systems)--that result from the pulse broadening associated with the material dispersion of the glass fiber. It is also desirable in wavelength division multiplexing as it makes possible a reduction in the spectral separations between the different wavelengths employed. Of course, single wavelength operation of a laser diode is desirable for other uses. The term "single wavelength" means a narrow spectral output corresponding to single mode operation.
"Single mode" operation, as originally used by those skilled in the art, meant that under CW operation the secondary modes were suppressed in intensity by a factor of at least 10 with respect to the primary mode. It is now well known that under high speed pulsed modulation, diodes that have single longitudinal mode operation under CW conditions may have an output with significant intensity in an unwanted longitudinal mode, i.e., the longitudinal mode may vary from pulse to pulse. Such diodes are not a solution to the problem of mode stability as they may lead to a significant error rate.
The origin of multiwavelength output from a semiconductor laser may be understood from the following discussion. In a semiconductor laser diode, a single Fabry-Perot cavity of length L is formed between two parallel and usually partially reflecting faces. The Fabry-Perot mode spacing, .DELTA..lambda., is approximately .lambda..sub.0.sup.2 /2N.sub.eff L, where .lambda..sub.0 and N.sub.eff are the wavelength of the lasing radiation in air and the effective refractive index of the propagating mode in the laser, respectively. Typical semiconductor laser diodes operate with at least several longitudinal modes because the decrease of the gain of the active medium near the peak of the gain spectrum is not rapid enough to suppress stimulated emission into adjacent modes, i.e., modes other than the primary or most intense mode.
Consequently, the semiconductor lasers presently manufactured generally emit radiation at several wavelengths unless an effort is made to narrow the spectral output to a single mode, i.e., essentially a single wavelength or frequency because the width of the individual modes is extremely narrow. Width means full width at half maximum. Several approaches have been tried in attempts to narrow the spectral output to essentially a single wavelength. One approach to obtaining single wavelength operation from semiconductor laser diodes under CW or high speed modulation is the use of a coupled cavity configuration which can be either external or integrated with the light source.
Single longitudinal mode operation in coupled cavities may be understood by considering the Fabry-Perot mode spacing of the coupled cavity configuration. If the cavities are of significantly different lengths, the coupled mode spacing, .LAMBDA., is approximately equal to that of the shorter cavity and may be increased by further shortening that cavity. If the cavities are approximately equal in length, then: EQU .LAMBDA.=.DELTA..lambda..sub.1 .multidot..DELTA..lambda..sub.2 /.vertline..DELTA..lambda..sub.1 -.DELTA..lambda..sub.2 .vertline.(1)
or EQU .LAMBDA.=.lambda..sub.0.sup.2 /2.vertline.N.sub.eff1 L.sub.1 -N.sub.eff2 L.sub.2 .vertline. (2)
where the subscripts in both equations refer to the first and second cavities, respectively. Thus, the Fabry-Perot mode spacing of the coupled cavities can be increased, and unwanted longitudinal modes suppressed, by having cavities of slightly different lengths. It was generally believed by those attempting to obtain single longitudinal mode operation by use of a multicavity configuration that a critical wavelength matching condition existed which placed restrictions on the lengths of the cavities and the spacing between them so that at least one of the modes of the coupled cavity was located within the gain profile of the semiconductor laser. Satisfaction of such a condition was often believed by those skilled in the art to require precise cavity lengths and intercavity spacing.
Multicavity configurations have been described in the literature. U.S. Pat. No. 4,284,963, issued on Aug. 18, 1981 to Allen et al, described a laser diode that had a predominantly single longitudinal mode output although adjacent longitudinal modes had easily measurable intensity. The laser, which was electrically contacted by single electrodes on both the top and bottom surfaces, i.e., it was a two terminal device, was cleaved perpendicular to the active stripe to form an internal etalon in the cavity. While single longitudinal mode operation was described, the diode was operated under CW conditions and the additional, i.e., the unwanted, longitudinal modes were sufficiently intense to prohibit use of this laser diode in high data rate communications systems. This laser is also described in Proceedings of the Society of Photo-Optical Instrumentation Engineers, 157, pp. 110-117, 1978.
Coldren et al described an integrated optical device capable of single dominant longitudinal mode operation in Applied Physics Letters, pp. 315-317, March 1981. A high aspect ratio groove was formed in the device by reactive ion etching to form two optically coupling cavities. For single longitudinal mode operation at an arbitrary output level, Coldren et al calculated that cavity lengths within a prescribed regime were required. The intensity ratios disclosed for the primary to secondary longitudinal modes were relatively small and the two sections were electrically connected with a series resistance of approximately 0.5 ohms. Latter work by members of the same group described, for example, as published in Electronics Letters, 18, pp. 901-902, Oct. 14, 1982, better results for a device with relative lengths of the two cavities having a ratio of 8:1 and an intercavity spacing of approximately 1 .mu.m. Although a deep groove separated the cavities, there was a lower measurable intercavity resistance. However, the ratio of the primary to the suppressed secondary modes was higher than in the first paper.
It was apparently believed by those skilled in the art that the reactive ion or chemical etching was preferable to cleaving as this technique of forming the cavities guarantees that the active stripes are precisely aligned with respect to each other and that the cavities have the desired dimensions and are precisely spaced with respect to each other. Satisfaction of these conditions guarantees good optical coupling and, furthermore, does not require leaving the substrate. It was believed that the power transmitted or reflected by the surfaces coupling the cavities was a sensitive function of the intercavity spacing. See, for example, IEEE Journal of Quantum Electronics, QE-18, pp. 1679-1688, October 1982. Strong intercavity coupling permitted better control of the longitudinal modes. Further, it was believed by some that better output was obtained when one cavity was much longer than the other cavity. Cavities of approximately equal length produced less clean output.
Additional multisection optical devices formed by cleaving are disclosed in IEEE Journal of Quantum Electronics, QE-4, pp. 125-131, April 1968 (Kosonocky and Cornely) and IEEE Journal of Quantum Electronics, QE-16, pp. 997-1001, September 1980 (Chang and Garmire). Both devices described in these papers were formed by mounting the single chip on a substrate, and, after scribing, bending to form cleaved mirror surfaces on the diodes. Neither paper reported single longitudinal mode operation. This is understandable because the lasers used were broad area lasers. The active area of the, for example, Chang et al device had a width of approximately 50 .mu.m. Due to the filamentary nature of semiconductor lasing, i.e., the lasing occurs in filaments less than 10 .mu.m in transverse dimensions, single longitudinal mode operation was not obtained in such a broad area laser. Additionally, one of the diode surfaces used by Kosonocky et al was intentionally lapped at an angle with respect to the principal axis of the active layer to destroy the Fabry-Perot cavity. Thus, one of the diodes of the Kosonocky et al device never lased.
Other multisection semiconductor laser devices have been described in the literature. For example, U.S. Pat. No. 3,303,431, issued on Feb. 7, 1967 to Fowler, described a broad area laser device suitable for some optical logic circuits. Two broad area semiconductor lasers were aligned end to end, and while the current density threshold was lowered, i.e., the threshold for the laser pair was lower than the sum of the individual laser thresholds, there was reported to be an absence of enhanced mode selectivity. In fact, the output spectrum was predominantly that of one laser with the emission from that laser being amplified by the second laser. Additionally, the only logic operation the device could perform was an AND gate whose operation relied on the observation that the threshold for the laser pair was lower than the sum of the individual laser thresholds.
Several multicavity injection locking devices have been described. U.S. Pat. No. 3,999,146, issued on Dec. 21, 1976 to Lang et al, described yet another multisection semiconductor laser device. The initial optical spiking and relaxation oscillations of the output intensity associated with the pulsed electrical input were suppressed in the disclosed device by injecting a light beam of predetermined intensity and the desired wavelength from a first laser diode into a second semiconductor laser diode. Thus, the longitudinal mode selection believed necessary for injection locking was obtained externally to the second semiconductor laser diode which, according to this disclosure, should not have an effect on the operation of the first laser diode. U.S. Pat. No. 4,101,845, issued on July 18, 1978 to Russer, described both method and apparatus for suppressing relaxation oscillations in the optical pulses when the diode is under modulation by using two separately controllable semiconductor lasers. One laser operated under CW conditions and had a high spectral purity because of external or distributed feedback mechanisms while the other laser was pulse modulated and had poorer spectral purity. The term "spectral purity" refers to the longitudinal mode spectrum with higher purity meaning fewer modes are present at some intensity level. The lasers were desirably optically isolated from each other so that the second laser had no effect on the behavior of the first laser although the reverse was obviously not true. Fekete et al described an integrated device in Applied Physics Letters, 37, pp. 975-978, Dec. 1, 1980, which was also an injection locking scheme. The device had a branched waveguide, which also functioned as a light source, with the light going into the laser section. The device was integrated and suppressed the relaxation oscillations of the light output when the device was switched. The essential principle of all injection locking devices is that the injected light lowers the optical amplitude at the beginning of the lasing pulse and increases the damping of the laser output thus suppressing the relaxation oscillations.
Multisection optical devices have been fabricated for still other purposes such as frequency modulation. For example, an article by F. K. Reinhart and R. A. Logan in Applied Physics Letters, 36, pp. 954-957, June 15, 1980, described a AlGaAs integrated optical circuit using an electro-optic frequency modulator. The diodes were broad area and the multi-layer structure was epitaxially grown on a (110) oriented GaAs substrate rather than the more commonly used (100) oriented substrate. This permitted the appropriate electro-optic effect to be obtained for frequency modulation. The circuit comprised a laser and an extra-cavity detector which were optically coupled to each other by a passive semiconductor waveguide. The laser and waveguide were coupled by the evanescent electromagnetic field of the radiation in the active layer.